Just think of it, a sound-sampling drum machine that also lets you synthesise percussion sounds and stores voice and pattern data on floppy disk. We preview Britain's best MIDI bet yet - and talk to the men behind it.

You probably don't need me to tell you that it takes a good deal of nerve to enter into the commercial arena of hi-tech musical instruments, especially if you're an unknown name and attempting R&D on a shoestring budget. Yet that's exactly what a new British company, Anvil Synthesisers, have tried to achieve with their advanced digital drum machine, the Anvil.

The Anvil story actually starts as far back as three years ago, but surprisingly for a project as complex as this, only a couple of people have been directly involved in the development. The men in question are musician and producer Rod Bowkett, who acted as the main creative drive, and engineer Roy Gwinn, who supplied the necessary hardware and software expertise.

If it's possible to sum up what the Anvil's designers have set out to achieve, it's to produce something in one box that'll do all the things people are using half-a-dozen different boxes for. By that, I mean everything you'd expect from drum synthesis, sampling drum machines, electronic drum modules, and so on.

Thus, the Anvil as it will be appearing in the near future is not only a sampling drum machine in the traditional sense, but also a percussion synthesiser that can be played from pads via MIDI. Furthermore, all its sounds can be stored on disk to be loaded up into RAM as and when you want to play, mix, or modify them.

Inside and Out

To begin with, the Anvil looks like no other programmable drum machine currently available. So it comes as little surprise to learn that its external layout (which has now earned the attentions of Design magazine, among others) came about as a result of collaboration with London's Royal College of Art.

It seems the College has a scheme whereby students work on projects that actually see the light of day, rather than remain within the academic seclusion of College galleries. A friend of a friend made the necessary introductions, and Hey Presto, the Anvil emerged as one of the RCA's 1984 projects.

The College team's intention, according to Bowkett, was to make the machine look thinner than it really was, and as black as possible. Well, looks like they succeeded on both counts...

Let's take a look inside that attention-getting exterior. The top PCB holds a Z80 processor and its attendant chips, and serves the sequencer and the control panel functions. There's an eight-bit parallel communication bus to the 8086-based bottom board with MIDI connections at the rear, plus a tape interface planned originally for saving of sequences and sounds, but cancelled very early on in the project. In fact, the tape interface is currently being used for SMPTE code purposes, but the design team are seriously considering replacing this with a custom SMPTE chip that'll cope better with varispeed.

Few modern musicians or engineers will let you forget the importance of informative displays on today's hi-tech instruments. After all, the more complex a machine gets internally, the more information the average user needs to exploit that complexity to the full. The Anvil's display is a 40-character indicator panel, with a 128-character set that incorporates special symbols for such things as notes. As it happens, the bar graph used in the machine's Sample mode is also built up from said special characters.

Users accustomed to working in ill-lit studios and concert venues (hands up the engineer who isn't) can take heart from the fact that the Anvil's main display unit is a specially-chosen fluorescent job, and the same reasoning lies behind the use of a number of electroluminescent areas - the sort of thing you'd normally find on car dashboards - for lighting up control functions above and below the panel's slider controls. Only problem for the designers was getting hold of an inverter that could power the displays, but that's another story.

The next thing to mention is the control strategy used for the Anvil's membrane pushbuttons. Interestingly, these buttons only illuminate if pressing them would cause something to happen, a condition that varies depending on what operational mode you're in at the time. By way of explanation, pushing the panel's Shift switch shows you precisely which buttons will light up in any given mode.

Anyway, now seems as good a time as any to list the various modes that exist within the Anvil's control system. In no particular order, these are Synth, Sample, Process, Add, Real-Time Parameter (RTP), Sequencer, Delete, and Menu.

As we mentioned in last month's Rumblings, the Anvil has 16 channels all told, though it seems likely these'll be most commonly used as eight stereo pairs. Their arrangement is pretty versatile; the only thing you can't do is play out the A and B sides of each channel pair at different pitches at the same time. That's because although there are 16 channels of DMA on the 8086 board for sending sound channels to the DACs, there are only eight counter-timers to actually generate the pitches. So, if you set the machine up to play, say, Channels 1A and 1B at different pitches, it'll do so quite happily providing they're not both happening at the same time. If a pitch conflict does exist, the channel that started earlier is terminated, and the pitch is changed.

"The Anvil's drum synth waveforms are built from tables of sinewaves, so you can perform additive synthesis before the sound gets to the filter."

That limitation might seem serious at first, but given that 16 simultaneous voices are a pretty unlikely event in any 'conventional' drum sequence, it shouldn't prove too crucial a failing in practice.

Synthesis

Entering Synth mode presents you with a layout very much like that of a normal analogue synth, with an oscillator, filter, envelope generators, and so on (see accompanying block diagram). First, there are controls for setting the pitch in quarter-tones over a 10.6-octave range, with a fine-tune control thrown in for good measure. The 'oscillator' is a bit unusual, in that it doesn't just have the usual four waveshapes and noise source that are then processed by subtractive synthesis in the vast majority of today's (and indeed yesterday's) analogue synthesisers. Instead, the Anvil's waveforms are built from tables of sinewaves, which means you're able to perform additive synthesis before the sound even gets to the filter. Hence the odd parameters that are attached to the oscillator, like 'harmonic amplitude', 'harmonic symmetry', and 'odd/even crossfade'. What's more, all these parameters can be changed dynamically by routing the appropriate ADLR to them.

Next in the synthesis line is the filter, which is basically a four-pole type with variable resonance, centre frequency, filter mode, and so on. It also has a filter offset: make this value anything other than zero, and it introduces two more filters at a specific displacement above the centre frequency. And since all this is done in software, it doesn't cost anything other than processing time to add more filters, which has got to be good news for all those recession-bitten recording studios out there.

As I've already implied, the envelope generators (there are five of them) are all of the ADLR variety, the 'L' standing for Lag time. Additionally, you can route any of the EGs to just about any conceivable destination within the synth section, and there's a variable delay time right at the beginning of the cycle. Finally, pushing the Play button tells the processor to synthesise the finished sound into memory, and a countdown character at the end of the display tells you how far it's got.

But with so much in the Anvil's sound synthesis department taking place in software, the time it takes for all the sound-modifying processes to actually happen is of crucial importance. At the moment, the designers estimate that the ratio between processing time and the actual length of the sound is about 10:1, which I reckon could be a little prohibitive.

In fact, the synth software also extends into Process mode, as this allows you to process any sound in memory with the same sort of filter and ADLR set-up, once you've defined your start and end points for a sample. Don't panic, you haven't missed anything - I'll be going into the sampling side of things a little later on.

Next we come to Add mode, which allows you to add various memory locations together into one; RTP mode, which is the way in which the single parameter that's recorded in a sequence - touch, pitch, pressure, or whatever - is used to modify the sound in a given channel; and Menu mode, which is basically everything else, including the disk functions for loading and saving sounds and sequences. I'm assured that when the disk software is completed, you should be able to load up all 13 seconds' worth of sounds on each 800K, 3.5" disk in about 40 seconds. Which, if my arithmetic serves me correctly, works out as something like a couple of seconds for a decent-length drum sound.

System Philosophy

Now, the facility the Anvil has for simulating an analogue synth purely by number-crunching into memory is certainly a first for any sort of sampling system - percussion or otherwise. But the reasons for doing that aren't simply to simulate an analogue synth. Gwinn and Bowkett are being a good bit more ambitious than that, as the latter explained to me: 'the synthesis potential is actually far more flexible than that of any analogue synth I know of. For instance, rather than just having fixed filter modes like bandpass, lowpass, and highpass, the Anvil makes it possible to program a cross-fade between any of them, thereby giving a vast range of potential filter characteristics. And I certainly can't conceive of an analogue synth that's capable of layering together an infinite number of the results that sort of synthesis can provide.'

That's where the Add mode comes in. With it, you can synthesise eight different sounds, mix them together, and then use that sum as the basis for adding more on top. The sliders on the front panel allow you to set the levels of each of the eight channels in the mix, and pressing Play records the mix into memory. Then you can go back and delete the individual sounds that were previously in memory and load up some new ones in their place.

The great thing about this is that if you want to add a touch more click to a bass drum, for example, all you have to do is synthesise the click in Synth mode, add this to the pre-existing bass drum sample in Add mode, and then save this on disk as the new, clickier bass drum. And, of course, the Add mode's analogue mixer means you can make such changes in real time.

The Anvil team are envisaging the starting up of disk libraries consisting of bits of sounds that can then be dropped on top of other sounds, as well as complete sets of sounds. But getting on to some rather more involved production of sounds brings us back to the speed with which you can hear changes in sound taking place in Synth mode. I asked Roy Gwinn if he foresaw any difficulties in that direction.

"The designers have set out to produce everything you'd expect from drum synthesis, sampling drum machines and electronic drum modules in one box."

'Hopefully not! If you try playing a sound before it's finished processing, it'll play as far as it's got at that particular time, which helps a lot to lessen the problem of the 10:1 processing time. The processing takes place at a rate of 24 microseconds per sample, and you can't issue very many instructions in that amount of time, even with an 8086 running at 8MHz. It'd be nice to make the synthesis real time, but you'd need a whole lot more hardware. Still, we are thinking of adding an 8087 arithmetic coprocessor (A cool £100 per chip - Ed) to the bottom board to help things along a bit.'

Sampling + Sequencing

Away from the intricacies of percussion synthesis and on to the oh-so-fashionable pastime of drum sampling. You can sample a sound into the Anvil's memory using either a microphone or line input, and there's a Level control at the back of the unit, this being allied to a bar graph display to give a visual indication of levels. Technically, sampling is 12-bit linear into 800K of RAM, with a sampling rate of 41 kHz. The six-pole input filter has a centre frequency of 18kHz, and on the output, there are sixteen separate five-pole filters with a -3dB cutoff at 16kHz.

If you're not particularly au fait with the jargon that currently engulfs the sampling process, I'll just say those figures make very impressive reading - and listening. In fact, the sample rate and filter parameters may yet be better still, if time and cheap parts suppliers permit.

As for the sequencer that links the Anvil's synth and sampling sections together to form programmable rhythm patterns, this is actually rather similar in spec to that fitted to the original Linn, though it has a tree structure comprising Links, Chains, and Songs. Sadly, the machine lacks any form of step-time facility at present, but there is a variable correction factor which quantises notes as they're keyed in in real time. The RTP value is entered either from the RTP slider as you're pressing the buttons to get the triggers in, or, if you're playing in a sequence from the MIDI, direct from the key value being played.

Moving on to the aforementioned tree structure, the length of a Link can be defined by pressing a couple of keys whilst listening to the metronome, or you can set up a time signature in the time-honoured fashion. You can build up to 99 individual Links, chain those together into up to 99 Chains, and then define a Song out of those. Only one Song exists in memory at any one time, not because it uses up a lot of memory, but simply because you're encouraged to use the disk for saving and loading sequences!

A built-in external triggering facility allows you to enter data in real time (from an electronic drum pad, for example) straight into the sequencer and then either process it or use it as it is. That way, you can combine quantised entry from the keypad on the Anvil's front panel with real-time entry from pads, all within the same Link if that's what you want.

This section is far from being every-man's answer to drum sequencing, however. There's no equivalent to the Linn's repeat trigger facility, and there isn't even any kind of flam option, even though machines a tenth of the Anvil's cost offer just such a provision. Actually, Bowkett and Gwinn would like to see both features built into their baby before too long, even though it seems unlikely that initial production Anvils will be so-equipped.

But my main source of disappointment with the Anvil's sequencer is that quantisation can only be applied while you're recording - not during playback. Bowkett defends the omission stoutly. The thing you often need to do on a drum machine is to have different quantisation levels within one measure, but simply being able to apply quantisation during playback wouldn't allow you to do that, so I can't really see the point. I certainly agree that keyboard sequencers need to have quantisation that's applied after recording, but for drum sequencers, I think the Linn approach works fine. I guess if enough people ask for it, we'll fit some sort of step-time facility, but let's see what happens first. Certainly it'd be a bit messy doing that on the Anvil's display, though it is a better display than on most drum machines - look at the Linn 9000, for instance. There are ways it could be done, but there's certainly no way 16 simultaneous events could be displayed.'

But given the Anvil's breadth of features, isn't it likely to be used as much more than just a standard drum machine? Perhaps users will find themselves needing a sequencer that's more a combination of both keyboard and drum sequencers.

'Yes, that's certainly possible. But one way of achieving that is to program in step-time using a MIDI sequencing package on a home micro. I think that's better than trying to force the Anvil's 40-character display into doing more than it's designed for.'

Conclusions

It's a strange paradox that the further up the technological ladder you climb, the less chance you have of pleasing all the people most of the time. It's a hard life, trying to get the cost-vs-facilities balance exactly right for as big a range of potential users as possible, yet I reckon Bowkett and Gwinn have done a better job than most, certainly in the upmarket drum machine stakes.

The lack of a step-time programming facility is the only serious omission I can find, and that's far outweighed by the sheer degree of imagination the Anvil team have shown in implementing the facilities their machine does incorporate.

The scope of the percussion synthesis section is extraordinary, the arrangement of channels is both neat and well thought-out, and the sampling is as good as any you'll find in this market area.

The only problem is that, as so often happens in circumstances like these, the people behind Anvil have grossly underestimated the time - and the money - it would take to get their project off the ground. As a result, the first version of the Anvil will probably be minus the synthesiser side, which still needs a fair bit in the way of software-writing. However, it will have the disk drive, sampling, sequencer, and MIDI as standard, and the company are thinking in terms of August's British Music Fair for the first public demonstration of this version. Software updates will take care of the synthesiser and SMPTE departments as and when work is completed.

As far as the price is concerned, Anvil are intending to sell the unit for around £5000. That might sound a lot for a drum machine no matter how extensive its facilities, but bear in mind that, with a few hardware modifications and some rather more extensive software rewriting, the Anvil could form the basis of a MIDI-compatible, polyphonic sound-sampling keyboard and digital recorder. And that isn't just wishful thinking on the designers' part...

For more information on the Anvil Percussion Synthesiser, contact Anvil Synthesisers Ltd on (Contact Details).